APRA Insurer Statistics XLSX: What the Regulator's Workbooks Contain

The Australian general insurance market, sized by its own supervisor. The Australian Prudential Regulation Authority receives statutory returns from every authorised general insurer in the country and publishes the aggregate result as statistical workbooks - gross and net earned premium, claims incurred, underwriting result, investment holdings and capital adequacy, quarter after quarter, reaching back to December 2002. APRA insurer statistics arrive through three connected products: the Statistics Portal, the discovery index announcing every release across all five APRA-regulated industries; the Quarterly General Insurance Performance Statistics, the aggregate performance series itself; and the National Claims and Policies Database (NCPD), which descends from market aggregates to transaction-level policy and claim records.

That trio is the only regulator-published record among the eight primary datasets in the reinsurance slice carrying Australian carrier-level depth, and it scores 8/10 on Datadory's rubric against a catalogue average of 7.81 - a mark shared by just 417 of 1,744 catalogued datasets. Australia-centred records are genuinely scarce: 8 of 1,744. Everything below reads those three products as one dataset, because that is how they behave when delivered through one schema.

Three products, three shapes - all landing on the same typed spine:

# Product 1 - QUARTERLY GENERAL INSURANCE PERFORMANCE STATISTICS
grain    : industry/class-of-business/state x quarter
fields   : Gross Earned Premium | Net Earned Premium | Claims Incurred
           | Underwriting Result | Investment Holdings | Capital Adequacy
current  : September 2023 to March 2026 window, refreshed each quarter
archive  : December 2002 to June 2023, incl. institution-level database
           September 2017 to June 2023 at individual authorised insurers

# Product 2 - NATIONAL CLAIMS AND POLICIES DATABASE (NCPD)
grain    : year x product x class x ANZSIC industry/occupation
           x deductible band x state x limit-of-indemnity band
fields   : Product Type | Premium | Risk Count | Deductible Band | Limit Band
lines    : Professional indemnity + public/product liability (core);
           cyber and management liability standalone
history  : policies and claims underwritten since 2003, latest reference
           period December 2024

# Product 3 - STATISTICS PORTAL INDEX
grain    : one row per statistical publication
defaults : title | industry | document_type | published_date
span     : 75+ publications, five regulated industries, releases 2015-2025

Read the first shape and market structure falls out as arithmetic: rank classes of business by net premium and the concentration question answers itself before a model is fitted. Read the second and pricing arrives beside exposure - deductible bands and limit bands keyed to the same class codes, which is exactly what casualty severity curves and reinsurance recovery models consume. Read the third and release monitoring becomes a diff operation instead of a calendar reminder.

These shapes are standing anatomy verified during the August 2026 research pass; the values populate against the quarters and reference periods you name. Slots depending on your scope fill in the sample, so nothing is padded onto every delivery with guessed columns.

Depth outranks freshness here, and this collection has both directions covered.

• Temporal: the current performance series covers September 2023 to March 2026, but the underlying databases run to December 2002 - more than two decades of quarterly and annual history. Institution-level granularity concentrates in the archive: the institution-level historical database runs September 2017 to June 2023, so analysts needing entity detail work in the pre-2023 material rather than the current aggregates. The NCPD reaches further still, covering policies and claims underwritten since 2003 through the December 2024 reference period.
• Geography: national figures with state and territory cuts throughout - general insurance results slice by state, and NCPD cells carry a state dimension too, useful because catastrophe exposure and liability dockets concentrate unevenly across jurisdictions.
• Granularity: industry-class-state cells in the current release, individual authorised insurers in the archive, and policy-and-claim cells in the NCPD - the widest grain ladder any single regulator in the slice offers.

One seam cuts through the series. Archived figures were collected on reporting standards superseded on 1 July 2023, so pre- and post-2023 panels are not spliceable - the discontinuity is a collection-basis change, not a revision. Pin whichever basis your model assumes and record it beside the delivery date; the specifications accompanying each release remain the authority for column definitions across the boundary.

Two grains, two different questions. The quarterly performance tables answer how is the market performing: premiums written, claims incurred, underwriting profit, capital strength - sliced by class of business and state. The NCPD answers what is inside the book: which industries buy which liability lines, at what deductibles and limits, with what premium and risk counts behind them.

Teams that skip the routing step routinely rebuild, by hand, an answer the neighboring product already printed. Treat the table below as the router - and when the choice is genuinely contested, our head-to-head works through it cell by cell in NCPD vs quarterly performance statistics.

• Masking is a feature, not a gap. In NCPD extracts, asterisked cells indicate masking applied to protect confidentiality, not failed collection. Because masking occurs at different aggregation levels per product line, totals across the report files need not reconcile - read cells, not cross-footings.
• Specifications are the schema authority. Each release ships alongside a specifications workbook defining every data element. It is the arbiter of column names, particularly around the July 2023 reporting-standard boundary where definitions moved underneath stable-looking labels.
• Statutory is not GAAP. These figures are filed on prudential conventions; reconciling them to the GAAP revenue in a listed insurer's earnings release produces nonsense. Treat them as regulatory truth about the whole authorised population, not a second copy of any single company's accounts.
• Revisions happen inside the rolling window. Quarters within the current database window can be revised between releases, so provenance per pull - publication date alongside the values themselves - belongs in the warehouse, not in a folder name.

For pipeline-level treatment of reshaping regulator spreadsheets into model-ready panels, see data scientists use cases.

API, files, or your warehouse. Daily, weekly, or hourly.

Spreadsheet-native regulators are the norm, not the exception - roughly 410 of the 1,744 datasets Datadory catalogs ship XLSX against only 86 shipping Parquet - so conversion is handled upstream of your warehouse, not inside it. Request a sample first: name the classes of business, the states and the vintages, and the extract arrives shaped identically to the standing feed, so anything prototyped on it survives into production unchanged.

Ranked by how directly the collection answers their day job:

1. Actuaries & pricing teams anchor liability severity and deductible-band analysis in NCPD cells, then benchmark the resulting book against industry underwriting margins from the quarterly tables - regulator-collected denominators no carrier's own portfolio view can supply.
2. Investors & quants read Australian insurer strength off underwriting results and capital adequacy rather than press releases, with institution-level archives supporting entity-level screens back to September 2017. See investors & quants.
3. Reinsurance brokers & treaty underwriters size ceded exposure state by state and line by line, pairing catastrophe-loss context with the regulator's account of who writes what where.
4. Data scientists & ML engineers get a panel built for entity resolution across the July 2023 basis break, plus claim-level labels for severity models that aggregate tables cannot substitute for.
5. Journalists, academics & students cite the prudential supervisor's canonical compilation when an insurer's solvency or a market's profitability becomes a story; the method is worked through in citation-grade research.

Every publisher in the reinsurance slice chose differently, and format determines workflow more than anything else does. The table lines up the six records an analyst weighing Australian prudential exposure most often compares - read it as a routing decision, not a ranking. Where catastrophe exposure meets Australian balance sheets directly, NOAA vs the APRA Statistics Portal works through that pairing cell by cell.

Start with the APRA Statistics Portal dataset page for the full field dictionary and to request a sample shaped to your classes of business, states and vintages. For the whole pool of eight primary datasets, the reinsurance data hub holds the pooled view and best reinsurance datasets carries the ranked comparison. The reinsurance data guide maps the full slice, showing where the APRA workbooks sit beside Lloyd's syndicate accounts, IRDAI's Indian archive and OECD country indicators.